3 ms·

I see the same issue in optical flow. Everyone and their grandma are overfitting Sintel. But everyday phenomena, like changing light conditions or trees with m

I see the same issue in optical flow. Everyone and their grandma are overfitting Sintel. But everyday phenomena, like changing light conditions or trees with many small branches cause the SOTA to fail miserably. I even approached research teams with my new data set that they could use to measure generalization, but the response has been frosty or hostile. Nobody likes to confirm that their amazing new paper is worthless for real world usage.

Tade0 7y ago
This is one of the reasons why I never got myself to finish my master's thesis. My advisor advised me to use synthetic data, because the work I was supposed to build upon did as well. I insisted on real data which I generated myself and, unsurprisingly, it showed how undoable this whole thing was(using stereo audio data to localise speakers) with the suggested methods.